QIF-T0017
highTransfer learning backdoor
Tier 3 — Demonstrated (Lab-proven)
Legacy status: DEMONSTRATED
Backdoor persists through transfer learning from pre-trained BCI models. Attacker poisons upstream model, downstream users inherit the backdoor. Common in EEG foundation model paradigm.
Technique Details
- Tactic
- QIF-M.SV
- Status
- DEMONSTRATED
- Bands
- S2
✚ Therapeutic Application
Backdoor propagation via transfer learning from compromised pre-trained BCI model
Neural Impact
1 of 7 neural bands affected
Drag to rotate. Click a region to learn more.
Click or hover over a glowing region to see the attack techniques targeting it and their severity.
Scoring
NISS:1.1/BI:L/CR:H/CD:H/CV:I/RV:P/NP:T CVSS:4.0/AV:N/AC:L/AT:P/PR:L/UI:N/VC:L/VI:H/VA:L/SC:L/SI:H/SA:L Governance
Neurorights at Risk
This technique threatens 5 of the 4 proposed neurorights (Ienca & Andorno, 2017).
FDORA §3305 Compliance
- ! CVSS partially captures risk; neural dimensions missing
- ! Consent complexity under-matches neural impact (CCI/NISS mismatch)
Population Vulnerability
CRB vulnerability adjustment (γ=0.30) accounts for age, diagnosis severity, consent capacity, and device dependency.
| Population | NISS Base | Adjusted | Severity | Delta |
|---|---|---|---|---|
| Adult (Default) | 6.0 | 6.0 | Medium | - |
| Child (10yr) + ADHD | 6.0 | 7.1 | High ▲ | +1.06 |
| Adult with ALS | 6.0 | 7.0 | Medium | +0.97 |
Validation Status
Theoretical / Not yet validated. This technique has not been independently tested. See the validation dashboard for what has been tested.